. Author manuscript; available in PMC: 2025 Oct 29.

Published in final edited form as: J Chem Theory Comput. 2025 Oct 6;21(19):9710–9725. doi: 10.1021/acs.jctc.5c01193

Algorithm 1. Training Procedure of PBGNN for Approximating the PB Model.

Let $𝒟$ be the training dataset, consisting of $L$ molecules ${(ℳ^{(1)}, ϕ_{r x n}^{(1)} (r)), \dots, (ℳ^{(L)}, ϕ_{r x n}^{(L)} (r))}$ .

Let $Θ = {W_{h}, W_{f}, W_{o}}$ be the set of trainable parameters of PBGNN.